>benchmarks
kek

>>108630010
Big benis if true.

>>108630064
How else would you evaluate an llm is better than an other ?

>>108631506
Human surveys of people actually using it. Error benchmarks mean jack shit in the real world.

>>108630010
>numbers
waow line went up

>it'll come out before new years
>oh it's just delayed because of new years
>it'll come out in february
>it'll come out in march
>it'll come out in april

>we will never le accept external funding because we are le amazing hedge fund and will le fund ourselves
>pls gib 10bn usd

Has anyone used Deepseek R or v3 for 'vibe coding'? How does it compare to a model like Claude Sonnet 4.6?

>>108630010
China is unironically desperate if they keep putting off the actual release and just only throwing blatant fake news about it

>>108631506
Could start by not using benchmarks that have all the answers online.

>>108630010
mention taiwan is a country once and the rest of the conversation is no longer usable, no matter how hard you try to shift the topic back.